Question Regarding Bootstrap

Guy Hoelzer hoelzer at unr.edu
Tue Oct 19 15:01:44 EST 1999


In article <7ui2gn$i3v at net.bio.net>, James McInerney
<james.o.mcinerney at may.ie> wrote:

> > Second, the bootstrap values will not change for a maximum parsimony
> > analysis when you add invariable characters; although they can change a
> > little using some other methods, like maximum likelihood.  For maximum
> > likelihood, the result is not predictable; sometimes the bootstrap values
> > will increase, sometimes they will decrease.

[SNIP - thanks for the kind words]

> Bootstrap values can be affected by the inclusion or exclusion of
> constant characters for a parsimony analysis.

[SNIP]

You are correct and my statement was misleading.  When invariant sites are
included in the matrix, they can affect the resulting bootstrap values. 
Let me explain the source of my misleading comment.  It is smart (no cost)
to eliminate invariant characters from your matrix before estimating a
topology by maximum parsimony, because they add computing time without
having any influence on the MP topology.  However, there is a quiet debate
as to whether invariant sites ought to be included in MP boostrapping
analyses, because they can have a minor affect on the outcome.  I think
the invariant sites should not be included, because they carry no
phylogenetic information according to the MP criterion.  In effect,
including the invariant sites causes the sizes of bootstrapped matrices to
vary, which is contrary to the standard use of nonparamentric phylogenetic
bootstrapping.  One of the goals is to create a large number of equally
sized matrices.  Because I omit the invariant characters from MP bootstrap
analyses, they obviously have no effect.  However, this thinking caused me
to miss the point of the question.

BTW, I always include the invariant sites for distance and maximum
likelihood methods where they can affect the topology (or set of
topologies) that maximizes the tree estimation criterion.

-- 
Guy Hoelzer
Department of Biology
University of Nevada Reno
Reno, NV  89557




More information about the Mol-evol mailing list